ระดับและการดำเนินการตรวจสอบ

MongoDB gives you two independent knobs to tune schema validation behaviour: validationLevel controls which documents are subject to validation, while validationAction controls what happens when a document fails validation. Together they let you introduce validation gradually into existing collections without breaking current data or applications.

strict is the default validation level. In strict mode, every insert and every update must pass the validator—no exceptions. If an existing document in the collection already violates the schema, attempting to update it will still be checked against the validator. Strict mode gives you the strongest data quality guarantee but can be disruptive when applied to a collection with legacy non-conforming documents.

moderate level applies the validator only to new inserts and to updates of documents that already pass the validator. Existing non-conforming documents can still be updated without being forced to comply. This is a safe migration path: you enable the validator on a live collection without breaking updates to old documents that don't yet match the new schema.

Setting validationLevel to off disables validation entirely, even if a validator is attached to the collection. This is useful during bulk data migrations or emergency hotfixes where you need to write non-conforming data temporarily. Always re-enable validation after the migration window closes.

error is the default validation action. When a document fails validation, MongoDB rejects the write entirely and returns an error to the client. The document is not written. This is the safest setting for production because it prevents malformed data from ever entering the collection.

warn action allows the document to be written even if it fails validation, but logs a warning message to the MongoDB server log. This is useful during a transition period when you want to observe how many violations occur without blocking existing application traffic. After examining the logs and fixing the offenders, you can switch to error action.

The two settings compose independently. A common migration strategy is to start with validationLevel: 'moderate' and validationAction: 'warn'—new documents must comply but get logged on failure, while old documents are untouched. After observing the warnings and backfilling old data, switch to strict + error for full enforcement.

Check the current validation settings on a collection using db.getCollectionInfos(). The response includes options.validationLevel and options.validationAction alongside the full validator document. This is useful to audit all collections in a database and verify that production collections have strict enforcement enabled.

Certain database operations support a bypassDocumentValidation: true option that skips the validator for that specific write. Only users with the bypassDocumentValidation privilege can use this option. It is intended for trusted admin scripts and data migrations only—never use it in application code as it defeats the purpose of having a validator.

When validationAction is warn, each failed validation is written to the MongoDB server log with the WRITE component and WARNING severity. The log line includes the database, collection, the violating document's _id, and the specific schema rule that was broken. You can monitor these logs with Atlas's log viewer or by tailing mongod.log on self-hosted deployments.

Here is the recommended four-phase playbook for adding validation to an existing production collection:

• Phase 1: Set moderate + warn — no impact, observe violations
• Phase 2: Fix application code to send compliant documents
• Phase 3: Backfill old non-conforming documents with a migration script
• Phase 4: Switch to strict + error — full enforcement

In this lesson you learned: validationLevel (strict/moderate/off) controls which documents are validated, validationAction (error/warn) controls whether failures block or just log, and combining moderate + warn is the safest way to introduce validation on a live collection. Next up we learn how to evolve schemas without downtime on a running MongoDB cluster.